we recently posted a detailed tutorial for REMD in amber10,
you might want to look at that.
you are right that the GB models introduce helical bias,
much more so with HCT than OBC, which is more reasonable.
however, we also showed that igb-7 (neck GB) is probably biased
slightly against all secondary structures. as of now there is no "perfect"
choice, and igb of 5 or 7 (with the appropriate radii) is probably
the best choice. if the results depend strongly on the GB model between
these two, then there is likely not a strong conformational preference
anyway. ff99 is not good, I recommend ff99SB.
also keep on the watch for ion pairing, particularly with Arg,
which is also a known issue with many, if not most GB models.
keep in mind that simulations in explicit solvent may well take hundreds
of nanoseconds to converge, so differences between your results and
the explicit solvent MD need to be done only in the context of
well defined precision limits for data from the difference solvent models.
On Tue, Jul 8, 2008 at 12:24 PM, <guardiani.fi.infn.it> wrote:
> Dear Amber experts,
>
> I am currently performing REMD simulations in implicit solvent of
> a peptide whose coordinates were taken from a larger protein.
> My goal is to determine the equilibrium structure of the free
> peptide in physiologic conditions. The peptide is first minimized
> according to:
>
> Minimization
> &cntrl
> imin = 1,
> maxcyc = 500,
> ncyc = 250,
> ntb = 0,
> igb = 1,
> gbsa = 1,
> saltcon = 0.2,
> cut = 300.0,
> ntpr = 10,
> ntx = 1,
> /
>
> The minimized structure is then cloned into 12 replicas (12 corresponds
> to the square root of the number of atoms in my peptide). The target
> temperatures of the replicas are chosen in geometric progression and
> are: 310, 332, 355, 380, 406, 435, 465, 498, 533, 570, 610, 652.
> Each replica is heated from 0 K to the corresponding target temperature
> in this way:
>
> Equilibration
> &cntrl
> imin = 0,
> cut = 24.0,
> igb = 1,
> gbsa = 1,
> saltcon = 0.2,
> nstlim = 500000,
> dt = 0.001,
> ntt = 3, gamma_ln = 1.0,
> tempi = 0.0, temp0 = 310.0,
> ntx = 1, irest = 0, ntb = 0,
> nscm = 50,
> ntpr = 100, ntwr = 100, ntwx = 100
> /
>
>
> I then perform 2.5 ns of REMD simulation that will be discarded in the
> computation of energies and other properties of interest. The input
> file for this part of the simulation is like this:
>
> MD run
> &cntrl
> imin = 0,
> cut = 24.0,
> igb = 1,
> gbsa = 1,
> saltcon = 0.2,
> nstlim = 2500,
> numexchg = 800,
> dt = 0.001,
> ntt = 3, gamma_ln = 1.0,
> tempi = 310.0, temp0 = 310.0,
> ntx = 5, irest = 1, ntb = 0,
> nscm = 50,
> repcrd = 0,
> ntpr = 500, ntwr = 500, ntwx = 500
> /
>
>
> After that I perform other 30 ns of simulation through 3 restarts with
> input files of this kind:
>
>
> MD run
> &cntrl
> imin = 0,
> cut = 24.0,
> igb = 1,
> gbsa = 1,
> saltcon = 0.2,
> nstlim = 2500,
> numexchg = 4000,
> dt = 0.001,
> ntt = 3, gamma_ln = 1.0,
> tempi = 310.0, temp0 = 310.0,
> ntx = 5, irest = 1, ntb = 0,
> nscm = 50,
> repcrd = 0,
> ntpr = 500, ntwr = 500, ntwx = 500
> /
>
> My problem is that the equilibrium population is completely dominated by
> helical structures, which is in disagreement with the simulation of some
> colleague who is performin explicit-solvent simulations on the same peptide.
>
> Basically I think there are two main sources of error:
>
> 1) I used the ff99 force field without noticing that it introduces a bias
> toward helical structures. I am now not sure which force field is more
> convenient for my simulations. Do you recommend I should use the ff03
> force field or rather the ff99SB force field, or maybe yet another one ?
>
> 2) I performed my simulations with the Hawkins, Cramer, Trular generalized
> Born model (igb=1). However I have read a paper
> (J. Chem. Theory Comput. Vol 3, No 1, pag 156-169, 2007)
> where the authors report on a comparison of REMD simulations on Ala10
> and they conclude that OBC-GB and HCT-GB introduce a strong bias for
> alpha-helical conformations while the GBn method (igb=7) produces a
> conformational population in better agreement with experimental data.
> Do you think the GBn method is the best choice of GB model for my
> simulations ? And which set of atomic radii (PBradii) do you recommend ?
>
>
> I thank you very much for your help and I am looking forward to receiving
> feedback from you.
>
> Best regards,
>
> Carlo Guardiani
>
>
> PS: If you notice any mistake in the REMD protocol that I am using, please
> tell me.
>
>
> ----------------------------------------------------------------
> This message was sent using IMP, the Internet Messaging Program.
>
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
> to majordomo.scripps.edu
>
--
===================================================================
Carlos L. Simmerling, Ph.D.
Associate Professor Phone: (631) 632-1336
Center for Structural Biology Fax: (631) 632-1555
CMM Bldg, Room G80
Stony Brook University E-mail: carlos.simmerling.gmail.com
Stony Brook, NY 11794-5115 Web: http://comp.chem.sunysb.edu
===================================================================
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
to majordomo.scripps.edu
Received on Wed Jul 09 2008 - 06:07:47 PDT
